Abstract: Background: Early screening for autism spectrum disorder (ASD) has become a consensus, but it is unclear whether the children's health physical examination system can be used for screening, and the screening tools and paths need to be explored further. Objective: To explore the path and tools of ASD early screening suitable for children's health physical examination system. Design: Diagnostic accuracy study. Methods: From August 1, 2018 to October 31, 2020,the ASD screening clinical pathway was embedded in the child health physical examination system of Shanghai Changning Maternal and Child Health Hospital for children aged 1836 months. The gold standard was the Autism Diagnostic Observation Scale version 2 (ADOS2) combined with Diagnostic and Statistical Manual of Mental Disorders version 5(DSM5), and the modified checklist for autism in toddlers (MCHATR) + MCHATR/F or two behavioral observations (customized) were the test criteria. After primary MCHATR screening, positive children were rescreened by MCHATR/F, and negative children were rescreened by two rounds of behavior observation. Children with positive rescreening were diagnosed as ASD or not. Children with negative rescreening were inquried in routine physical examination, and after the diagnosis of ASD was confirmed, that would be recorded in the children health physical examination system. Those who were positive in the primary screening but were not rescreened and those who were positive in the rescreening but were not recorded in the children's health examination system were lost to followup and were not included in the analysis. Main outcome measures: The Youden's index of primary screening and secondary screening of ASD. Results: In the primary screening，there were 10,635 children, 5,579 boys with an average age of (22.3±2.9) months and 5,056 girls with an average age of (22.4±2.9) months. A total of 252 cases were positive and 38 cases were lost to followup (loss rate 15.1%). There were 53 cases positive in MCHATR/F rescreening, 13 cases were lost to followup (loss rate 24.5%), and 26 cases were diagnosed as ASD. A total of 161 cases were negative in MCHATR/F rescreening, and 4 cases were inquiny ASD by routine physical examination. Among 10,383 cases negative in the primary screening, there were 8 ASD in the 2018 subgroup, and 3 ASD in the 2019 subgroup by inquiry in the routine physical examination. In rescreening positive patients of the first behavior observation (n= 54),33 cases were positive, 18 cases were lost to followup (loss rate 54.5%), and 13 cases were diagnosed as ASD. There were 54 case of ASD dignosed by the overall screening, 41 boys and 13 girls with an average age of (26.2±5.9) and (23.8±5.3)months, respectively. The Youden's index of MCHATR was 0.539, the specificity was 98.4%, and the sensitivity was 55.6%. On the basis of positive MCHATR primary screening, the rescreening Youden's index of MCHATR/F was 0.78, the specificity was 91.3%, and the sensitivity was 86.7%. On the basis of negative MCHATR primary screening, the rescreening Youden's index of two rounds behavior observation was 0.81, the specificity was 99.96%, and the sensitivity was 81.2%. Eight children in the 2018 subgroup with negative results in MCHATR screening were diagnosed as ASD through inquiry in the following routine physical examination, and the age of diagnosis was (34.8±2.7) months. In 2019 subgroup of negative results in MCHATR screening, 3 cases were diagnosed as ASD during one round of behavior observation, and for negative subjects the age of diagnosis was(32.7±3.8) months. In the second round of behavior observation, 13 cases were diagnosed as ASD, and the age of diagnosis was (24.2±4.0) months. The difference between groups was statistically significant (F=22.809, P＜0.001), and there was also statistically significant difference in further pairwise comparison (P＜0.001, P=0.004). Conclusion: It is feasible for children aged 1836 months to embed ASD screening path in children's health physical examination system, and the screening efficiency is desirable. Behavior observation rescreening in children negative in the MCHATR screening can diagnose ASD 10 months earlier.

Key words: Autism spectrum disorder, Modified checklist for autism in toddlers, Screening, Behavior observation, Specificity, Sensitivity, Youden's index